[Streaming] Update PairDStreamFunctions.scala

[Jimmy-Newtron]

What changes were proposed in this pull request?

Added methods to retrieve keys or values of a pair DStream

How was this patch tested?

NONE

Added two methods that are available into the PairRDDFunctions.scala

[AmplabJenkins]

Can one of the admins verify this patch?

[srowen]

I don't think this worth new API methods since it's so easy to do this directly. Please see http://spark.apache.org/contributing.html too. I'd close this.

[Jimmy-Newtron]

I think that this new API methods is going to bring consistency between PairRDDFunctions and PairDStreamFunctions by providing the same API to access the same data.

I see the use case of a streaming application that is consuming a Kafka topic to retrieve the values as the simplest example of usage of this new API.

I accept the PR rejection, even if the new API is matching only positive aspects of the Code Review Criteria.

Positives:
Adds functionality needed by a large number of users
Simple, targeted
Negatives, Risks: NONE

Best regards

[rxin]

This is actually parity with PairRDDFunctions so I think it's OK to have them.

But again, having to add stuff like this is why we are doing structured streaming -- it's one single API for both batch and streaming.

[srowen]

I get it, it still seems not so worth it. The DStream API has always been mostly redundant with foreachRDD + the RDD API, and is itself maybe not something being actively updated anyway.